No abstract

The aesthetic defects of polymeric polarizers have a serious impact on the quality of thin film transistor liquid crystal display (TFT-LCD) panels. However, some of these slight and transparent defects can barely be imaged and characterized using conventional illumination. To inspect these special defects, a new and automated inspection method is proposed that employs structured-light illumination. A machine vision system that uses a LCD monitor to produce a binary stripe pattern was designed to enhance the imaging defects. Thus, subsequent image processing is straightforward and simple because of this enhancement. One hundred and fifty defect samples were successfully imaged and then processed and characterized by the Robust Principal Component Analysis (RPCA) algorithm. The total inspection accuracy was up to 99%. The experiments demonstrate that the width of the luminous stripe should be optimized at 2 to 3 times the size of the defect. The mechanism of the imaging enhancement is preliminarily discussed.

No abstract